User matching method and system based on social platform

Abstract

The invention provides a user matching method and system based on a social platform. The method comprises the steps that multi-dimensional user information data is acquired and preprocessed; feature extraction is carried out on the user information data to generate user feature data, and weights are given to the user feature data; constructing a user feature model for the user according to the weight of each piece of user feature data and a preset model training method; and calculating the similarity of the user feature models between the to-be-matched user and other users, and recommending other users corresponding to the similarity reaching a preset value to the to-be-matched user. According to the method, the multi-dimensional features of the users are comprehensively considered, the detailed and accurate user portraits are constructed, and matching recommendation is performed according to the similarity of the user feature models between the to-be-matched users, so that the limitation of matching only depending on single-dimensional information is avoided, the accuracy and reliability of user matching in a traditional social platform are greatly improved, and the user experience is improved. And the satisfaction degree and the participation degree of the user are obviously improved.

Description

Technical Field

[0001] The present application relates to the technical field of social network, and in particular to a user matching method and system based on a social platform. Background Art

[0002] With the rapid development of social networks, users have an increasingly strong demand to efficiently find social partners on social platforms who have interests that are highly consistent with their own, complementary personalities, and consistent values.

[0003] Traditional social matching methods are mostly limited to matching based on simple basic user information (such as age, gender, geographic location, etc.). This overly single model cannot fully and accurately reflect the diverse characteristics of users, which greatly reduces the accuracy of matching results and makes it difficult to improve user satisfaction. Summary of the invention

[0004] In view of this, this application proposes a user matching method and system based on a social platform, and the specific scheme is as follows:

[0005] In the first part, this application proposes a user matching method based on a social platform, which includes:

[0006] Acquire multi-dimensional user information data from one or more preset data sources, and pre-process the user information data;

[0007] Extracting features from the preprocessed user information data based on a preset feature extraction rule to generate user feature data, and assigning weights to the user feature data based on a preset weighting method;

[0008] Constructing a user feature model for the user according to the weight of each of the user feature data and a preset model training method;

[0009] The similarity of the user feature model between the user to be matched and other users is calculated, and other users corresponding to the similarity reaching a preset value are recommended to the user to be matched based on a preset recommendation strategy.

[0010] In a specific embodiment, the user information data includes: identity information data, and / or interest data, and / or online research data, and / or social platform interaction data.

[0011] In a specific embodiment, extracting features from the preprocessed user information data further includes:

[0012] The data type of the user information data is determined, and different feature extraction methods correspond to different data types of the user information data.

[0013] In a specific embodiment, the preset weighting method includes: a subjective weighting method, and / or an objective weighting method, and / or a combined weighting method.

[0014] In a specific embodiment, the “building a user feature model for the user according to the weight of each user feature data and a preset model training method” specifically includes:

[0015] Preliminarily constructing a user feature model for the user according to the weight of each of the user feature data;

[0016] Determining the existing social relationship data of the user, and using the social relationship data as a training sample for the preliminarily constructed user feature model;

[0017] The parameters of the user feature model are adjusted by optimizing an objective function based on the training samples to determine a final user feature model for the user.

[0018] In a specific embodiment, the “recommending other users corresponding to the similarity reaching a preset value to the user to be matched based on a preset recommendation strategy” further includes:

[0019] Sorting the similarities between the to-be-matched user and each of the other users;

[0020] The other users corresponding to the similarities that meet a preset threshold are sequentially formed into a recommendation list, and the recommendation list is sent to the user to be matched, wherein the recommendation list includes the user information data corresponding to each of the other users.

[0021] In a specific embodiment, it also includes:

[0022] Monitoring the user characteristic data of the user at a preset frequency, and if the user characteristic data changes, dynamically adjusting the weight of each user characteristic data and the user characteristic model for the user according to the change;

[0023] The similarity between the user characteristic model adjusted by the user and the user characteristic models of other users is recalculated, and other users corresponding to the similarities reaching a preset value are recommended to the user based on a preset recommendation strategy.

[0024] In a specific embodiment, it also includes:

[0025] One or any multiple of the user information data, the user feature data, and the user feature model are associated with the user and then stored.

[0026] In the second part, this application proposes a user matching system based on a social platform, including:

[0027] An acquisition module, used to acquire multi-dimensional user information data from one or more preset data sources, and pre-process the user information data;

[0028] An extraction module, used for performing feature extraction on the preprocessed user information data based on a preset feature extraction rule to generate user feature data, and assigning weights to the user feature data based on a preset weighting method;

[0029] A construction module, used to construct a user feature model for the user according to the weight of each user feature data and a preset model training method;

[0030] The recommendation module is used to calculate the similarity of the user feature model between the user to be matched and other users, and recommend other users corresponding to the similarity reaching a preset value to the user to be matched based on a preset recommendation strategy.

[0031] In a specific embodiment, extracting features from the preprocessed user information data further includes:

[0032] The determination module is used to determine the data type of the user information data, and the user information data of different data types correspond to different feature extraction methods.

[0033] Beneficial effects:

[0034] This application builds a detailed and accurate user portrait by comprehensively considering the user's identity information, interests, hobbies, social behavior and other multi-dimensional characteristics, and makes matching recommendations based on the similarity of the user feature models between the users to be matched. It avoids the limitation of relying solely on single-dimensional information for matching, greatly improves the accuracy and reliability of user matching in traditional social platforms, provides users with social matching recommendations that are more in line with their personalized needs, and significantly improves user satisfaction and engagement. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a flow chart of a user matching method based on a social platform according to an embodiment of the present application;

[0036] Figure 2 This is a schematic diagram of a user matching system module based on a social platform according to an embodiment of the present application;

[0037] Figure 3 This is an architecture diagram of a user matching method based on a social platform in an embodiment of the present application;

[0038] Figure 4 This is a user push flow chart for an embodiment of the present application.

[0039] Figure numerals: 1 - acquisition module; 2 - extraction module; 3 - construction module; 4 - recommendation module. DETAILED DESCRIPTION

[0040] Hereinafter, various embodiments disclosed in the present application will be described more fully. The present application may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments disclosed in the present application to the specific embodiments disclosed herein, but the present application should be understood to cover all adjustments, equivalents and / or alternatives falling within the spirit and scope of the various embodiments disclosed in the present application.

[0041] The terms used in the various embodiments disclosed in the present application are only used to describe the purpose of specific embodiments and are not intended to limit the various embodiments disclosed in the present application. As used herein, the singular form is intended to also include the plural form, unless the context clearly indicates otherwise. Unless otherwise specified, all terms used here (including technical terms and scientific terms) have the same meaning as the meaning generally understood by ordinary technicians in the field of various embodiments disclosed in the present application. The terms (such as the terms defined in the dictionary generally used) will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning, unless clearly defined in the various embodiments disclosed in the present application.

[0042] Example 1

[0043] Example 1 of the present application discloses a user matching method based on a social platform. This embodiment constructs a detailed and accurate user portrait by comprehensively considering the user's identity information, interests, hobbies, social behavior and other multi-dimensional features, and makes matching recommendations based on the similarity of the user feature models between the users to be matched, avoiding the limitation of matching based on only a single dimensional information, greatly improving the accuracy and reliability of user matching in traditional social platforms, providing users with social matching recommendations that are more in line with their personalized needs, and significantly improving user satisfaction and participation. The specific process is as follows: Figure 1 , Attachment Figure 3 and attached Figure 4 The specific plan is as follows:

[0044] A user matching method based on a social platform, as shown in the attached manual Figure 1 As shown, the method includes:

[0045] 101. Obtain multi-dimensional user information data from one or more preset data sources, and pre-process the user information data. In this embodiment, rich information about users can be widely collected from multiple data sources of the social platform. This information includes multi-dimensional feature data such as basic user information (such as age, gender, occupation, etc.), interests and hobbies (such as reading, sports, music, etc.), social behavior data (such as published content, comments, likes, etc.), and value orientation (obtained through questionnaires or specific behavior analysis). The collected data is then comprehensively cleaned, denoised, and normalized to ensure the accuracy and consistency of the data. For example, scientific filling methods (such as mean filling, median filling, etc.) are used for missing values ​​to accurately identify and correct outliers.

[0046] 102. Feature extraction is performed on the preprocessed user information data based on the preset feature extraction rules to generate user feature data, and weights are assigned to the user feature data based on the preset weighting method. In this embodiment, corresponding professional feature extraction methods are used for different types of user features. For example, for textual descriptions of interests and hobbies, they are converted into numerical vector representations with the help of word vector models in natural language processing technology (such as Word2Vec, GloVe, etc.); for social behavior data, relevant features are extracted by in-depth analysis of key factors such as user interaction frequency and interaction objects.

[0047] Then, the extracted features are quantified and each feature is given a scientific and reasonable weight. There are many effective methods to determine the weight, such as subjective weighting based on expert experience, objective weighting based on data analysis (such as principal component analysis, entropy weight method, etc.), or combined weighting method that combines the advantages of both, so as to accurately reflect the importance of each feature in social matching.

[0048] 103. Construct a user feature model for the user according to the weight of each user feature data and the preset model training method. In this embodiment, based on the quantized user features, a variety of model structures can be used to construct an advanced user feature model, such as a vector space model, a neural network model in deep learning (such as a multi-layer perceptron, a convolutional neural network, etc.). Taking the vector space model as an example, the multi-dimensional features of each user are represented as a feature vector, and each dimension of the vector corresponds to a different feature and its quantized value.

[0049] Furthermore, during the model training process, a large amount of labeled social relationship data (such as friend relationships, follow relationships, etc.) can be used as training samples, and the model parameters can be fine-tuned by optimizing the objective function (such as minimizing the error between the predicted relationship and the actual relationship), so that the model can accurately learn the intrinsic relationship between user characteristics and social matching.

[0050] 104. Calculate the similarity of the user feature model between the user to be matched and other users, and recommend other users corresponding to the similarity reaching the preset value to the user to be matched based on the preset recommendation strategy. In this embodiment, for the user pairs to be matched, the similarity between their feature vectors needs to be accurately calculated respectively. The similarity calculation method can use a variety of metrics such as cosine similarity, Euclidean distance, Pearson correlation coefficient, etc., and the appropriate method needs to be carefully selected according to the actual application scenario and data characteristics. For example, cosine similarity is suitable for measuring the similarity of vectors in direction, and the effect is significant when the feature distribution is relatively uniform; Euclidean distance focuses more on measuring the absolute distance of vectors in space, and is more sensitive to situations where the feature values ​​are more different.

[0051] Furthermore, based on the similarity calculation results, the user pairs are sorted in order, and the user pairs with higher similarity are selected as potential social matching objects. At the same time, in order to improve the diversity and personalization of recommendations, reasonable thresholds can be set to filter out user pairs with different similarity intervals for recommendation, effectively avoiding excessive concentration of recommendation results on a small number of highly similar users. Finally, the matching results can be presented to the user in the form of a clear and concise recommendation list, which can include detailed information such as the user's basic information, common interests and hobbies, matching scores, etc., so that users can fully understand and make wise choices.

[0052] In a specific embodiment, user information data includes: identity information data, and / or interest and hobby data, and / or online survey data, and / or social platform interaction data. In this embodiment, user information data covers multiple aspects, including but not limited to: detailed identity information data (such as age, gender, occupation, etc.), and interest and hobby data reflecting personal preferences (such as reading preferences, music tastes, etc.); in addition, it may also include online survey data in which users participate, which can reveal users' views and needs; at the same time, users' interaction data on social platforms (such as posts, comments, likes, etc.) is also an important source of information, which can reflect users' social behavior and influence.

[0053] In a specific embodiment, extracting features from the preprocessed user information data further includes:

[0054] The data type of the user information data is determined, and different feature extraction methods correspond to different data types of user information data.

[0055] In this embodiment, corresponding professional feature extraction methods are used for different types of user features. For example, for textual descriptions of interests and hobbies, word vector models (such as Word2Vec, GloVe, etc.) in natural language processing technology are used to convert them into numerical vector representations; for social behavior data, relevant features are extracted by deeply analyzing key factors such as user interaction frequency and interaction objects.

[0056] In a specific embodiment, the preset weighting method includes: a subjective weighting method, and / or an objective weighting method, and / or a combined weighting method.

[0057] In this embodiment, weights can be determined using a variety of effective methods, such as subjective weighting methods based on expert experience, objective weighting methods based on data analysis (such as principal component analysis, entropy weight method, etc.), or combined weighting methods that combine the advantages of both, so as to accurately reflect the importance of each feature in social matching.

[0058] In a specific embodiment, “building a user feature model for a user according to the weight of each user feature data and a preset model training method” specifically includes:

[0059] Preliminarily build a user feature model for each user based on the weight of each user feature data;

[0060] Determine the user's existing social relationship data and use the social relationship data as training samples for the initially constructed user feature model;

[0061] Based on the training samples, the parameters of the user feature model are adjusted by optimizing the objective function to determine the final user feature model for the user.

[0062] In this embodiment, the user's feature data is first read from the feature database to build an advanced user feature model. For example, when a neural network model is used, a deep learning framework is used to build a model structure, and the feature data is used as input for model training.

[0063] Furthermore, during the model training process, a large amount of labeled social relationship data (such as friend relationships, follow relationships, etc.) can be used as training samples, and the model parameters can be fine-tuned by optimizing the objective function (such as minimizing the error between the predicted relationship and the actual relationship), so that the model can accurately learn the intrinsic relationship between user characteristics and social matching. During the training process, optimization algorithms such as batch gradient descent and stochastic gradient descent are used to adjust the model parameters, with the goal of minimizing the loss function (such as the cross entropy loss function, the mean square error loss function, etc.), so that the model can accurately and deeply learn the relationship between user characteristics and social matching. Finally, the trained model parameters are stored in the model library for convenient call in the actual matching process.

[0064] In a specific embodiment, “recommending other users corresponding to a preset similarity value to the user to be matched based on a preset recommendation strategy” also includes:

[0065] Sort the similarities between the user to be matched and other users;

[0066] The other users corresponding to the similarities that meet the preset threshold are sequentially formed into a recommendation list, and the recommendation list is sent to the user to be matched. The recommendation list includes user information data corresponding to each other user.

[0067] In this embodiment, the system also includes: the system will accurately sort the similarities between the user to be matched and other users in order from high to low to ensure that the most similar users can be given priority. And screen out those users whose similarities meet the preset threshold. These users have a high degree of matching with the user to be matched, and are therefore considered as potential recommendation objects. Then, the system will arrange these other users who meet the conditions in order according to the previously arranged order to form a clear and organized recommendation list.

[0068] Finally, the recommendation list will be sent to the user to be matched by the system for review and selection. The recommendation list not only contains the basic information of other users, but also lists their corresponding user information data in detail, such as personal profile, interests and hobbies, so that the user to be matched can have a more comprehensive understanding of the characteristics of each recommended object and make a more satisfactory choice.

[0069] In a specific embodiment, it also includes:

[0070] Monitor the user's user feature data at a preset frequency. If the user feature data changes, dynamically adjust the weight of each user feature data and the user feature model for the user according to the change;

[0071] The similarity between the user characteristic model adjusted by the user and the user characteristic models of other users is recalculated, and other users corresponding to similarities reaching a preset value are recommended to the user based on a preset recommendation strategy.

[0072] In this embodiment, the system also includes: the system will continuously monitor the user feature data of each user at a preset frequency, such as daily or weekly. These feature data may cover multiple aspects such as the user's interests, behavior patterns, social activities, etc. Once the system detects that the user's feature data has changed significantly, such as the user has newly added certain interest tags or the behavior pattern has changed, the system will immediately respond and dynamically adjust the weight of each user's feature data. This means that according to the changes in user characteristics, the system will give new importance to different feature data to more accurately reflect the user's current status. At the same time, the system will also update the user's user feature model for the user to ensure that the model is consistent with the user's current feature data.

[0073] After completing the weight adjustment and model update, the system will recalculate the similarity between the adjusted user feature model of the user and the user feature models of all other users. This step is key to ensuring the accuracy of recommendations, because changes in user features may mean changes in their similarity with other users. Finally, based on the preset recommendation strategy, the system will recommend other users whose similarity reaches the preset value. This recommendation process not only takes into account the real-time nature of user feature data, but also ensures the accuracy and timeliness of recommendations by dynamically adjusting weights and models, thereby providing users with more personalized recommendation services that meet their current needs.

[0074] In a specific embodiment, it also includes:

[0075] One or more of the user information data, user feature data, and user feature models are associated with the user and stored. This associated storage method ensures the integrity and consistency of the information, allowing the system to quickly retrieve all relevant information about the user at any time based on the user ID or other unique identifier. In addition, this associated storage method also makes it easier for the system to update and maintain user information. When some user information changes, the system only needs to find the corresponding user record and then make corresponding modifications, without the need for large-scale search and update operations on the entire database, thereby greatly improving the system's operating efficiency and the accuracy of data processing.

[0076] The following is an example of a specific social matching process based on the method of this embodiment:

[0077] User A registered on the social platform and completed his personal information, including basic information (age 25, male, professional programmer, etc.), hobbies (like programming, reading science fiction novels, playing basketball, etc.), and filled out a questionnaire on values ​​(focusing on innovation, teamwork, integrity, etc.). At the same time, the system further enriches its feature data by deeply analyzing user A's historical social behavior (such as publishing programming technology articles, liking comments on science fiction novels, participating in basketball club activities, etc.).

[0078] The system collects relevant data of user A according to the data collection process, and converts the multi-dimensional features of user A into a feature vector through preprocessing, feature extraction and quantification, such as [0.8, 0.6, 0.4, 0.2, 0.7, 0.5, ...] (where each value corresponds to the quantified value of a different feature, assuming that the first few values ​​represent age, gender, occupation, programming interest, science fiction interest, basketball interest and other features).

[0079] When user B also completes a similar registration and information completion process, the system also constructs a feature vector for user B, such as [0.7, 0.5, 0.3, 0.6, 0.4, 0.3, ...].

[0080] When performing social matching, the system calculates the cosine similarity between the feature vectors of user A and user B. Assume that the calculated similarity is 0.85 (the similarity value ranges from 0 to 1, and the closer the value is to 1, the higher the similarity).

[0081] According to the preset similarity threshold (for example, 0.7), user A and user B are judged as potential matches. At the same time, the system comprehensively considers factors such as user A's preference for hobbies (such as a higher weight for programming interest) and the closeness of the social relationship between users (if two people have common friends or have participated in the same social activities, the matching degree will be appropriately increased) to optimize the matching results.

[0082] Finally, the system displays user B as a recommended object in user A's social matching recommendation list, and provides detailed information such as user B's basic information, common interests (programming, science fiction), and matching score (0.85), so that user A can fully understand and choose whether to interact with user B further.

[0083] Furthermore, in order to fully verify the effectiveness and superiority of the method of the present invention, the following rigorous experiments were conducted:

[0084] 1. Experimental Data

[0085] We collected real user data from a social platform, including basic user information, hobby tags, social behavior records, and existing social relationships (friendships, following relationships, etc.). We selected 10,000 users as experimental samples, of which 8,000 users were used for algorithm model training and 2,000 users were used for testing and evaluation.

[0086] 2. Comparison method

[0087] A comprehensive comparison is made with the traditional matching method based on basic user information (considering only age, gender, geographic location, etc.) and the matching method based on a single interest (matching only based on the main interests selected by the user).

[0088] 3. Evaluation Metrics

[0089] Precision, recall and F1 value are used as the main evaluation indicators. Precision is used to measure the accuracy of the recommendation, that is, the proportion of recommended matching objects that truly meet user needs; recall reflects the ability of the algorithm to discover potential matching objects, that is, the proportion of recommended objects that actually meet user needs; F1 value is the harmonic mean of precision and recall, which is used to comprehensively evaluate the performance of the algorithm.

[0090] In addition, a user satisfaction survey is introduced as a subjective evaluation indicator. A questionnaire is distributed to test users to ask them about their satisfaction with the recommendation results (very satisfied, satisfied, average, dissatisfied, very dissatisfied), and the satisfaction score is calculated.

[0091] 4. Experimental Results

[0092] After many scientific experiments and statistical analyses, the following valuable results were obtained:

[0093] The algorithm of the present invention is significantly superior to traditional matching methods in terms of accuracy, recall rate and F1 value. For example, compared with the matching method based on basic user information, the accuracy rate is increased by 35%, the recall rate is increased by 40%, and the F1 value is increased by 38%; compared with the matching method based on a single interest, the accuracy rate is increased by 28%, the recall rate is increased by 32%, and the F1 value is increased by 30%.

[0094] In terms of user satisfaction, the average satisfaction score obtained by the algorithm of the present invention is 4.2 points (full score 5 points), while the average satisfaction scores of the traditional methods are 2.8 points (based on basic user information) and 3.5 points (based on a single interest). This fully shows that users are more satisfied with the social matching recommendation results provided by the algorithm of the present invention, which further effectively verifies the effectiveness and practicality of the algorithm.

[0095] Example 2

[0096] Embodiment 2 of the present application discloses a user matching system based on a social platform, which systematizes a user matching method based on a social platform in Embodiment 1 to make it more practical. Figure 2 The specific plan is as follows:

[0097] This embodiment proposes a user matching system based on a social platform, including:

[0098] Acquisition module 1, used to acquire multi-dimensional user information data from one or more preset data sources and pre-process the user information data;

[0099] Extraction module 2, used for extracting features from the preprocessed user information data based on a preset feature extraction rule to generate user feature data, and assigning weights to the user feature data based on a preset weighting method;

[0100] Construction module 3, used to construct a user feature model for the user according to the weight of each user feature data and a preset model training method;

[0101] The recommendation module 4 is used to calculate the similarity of the user feature model between the user to be matched and other users, and recommend other users corresponding to the similarity reaching a preset value to the user to be matched based on a preset recommendation strategy.

[0102] In a specific embodiment, extracting features from the preprocessed user information data further includes:

[0103] The determination module is used to determine the data type of the user information data. User information data of different data types correspond to different feature extraction methods.

[0104] In a specific embodiment, the user information data includes: identity information data, and / or interest data, and / or online research data, and / or social platform interaction data.

[0105] In a specific embodiment, the preset weighting method includes: a subjective weighting method, and / or an objective weighting method, and / or a combined weighting method.

[0106] In a specific embodiment, “building a user feature model for a user according to the weight of each user feature data and a preset model training method” specifically includes:

[0107] Preliminarily build a user feature model for each user based on the weight of each user feature data;

[0108] Determine the user's existing social relationship data and use the social relationship data as training samples for the initially constructed user feature model;

[0109] Based on the training samples, the parameters of the user feature model are adjusted by optimizing the objective function to determine the final user feature model for the user.

[0110] In a specific embodiment, “recommending other users corresponding to a preset similarity value to the user to be matched based on a preset recommendation strategy” also includes:

[0111] Sort the similarities between the user to be matched and other users;

[0112] The other users corresponding to the similarities that meet the preset threshold are sequentially formed into a recommendation list, and the recommendation list is sent to the user to be matched. The recommendation list includes user information data corresponding to each other user.

[0113] In a specific embodiment, it also includes:

[0114] Monitor the user's user feature data at a preset frequency. If the user feature data changes, dynamically adjust the weight of each user feature data and the user feature model for the user according to the change;

[0115] The similarity between the user characteristic model adjusted by the user and the user characteristic models of other users is recalculated, and other users corresponding to similarities reaching a preset value are recommended to the user based on a preset recommendation strategy.

[0116] In a specific embodiment, it also includes:

[0117] One or more of the user information data, user feature data, and user feature models are associated with the user and stored.

[0118] This embodiment proposes a user matching system based on a social platform, which systematizes the user matching method based on a social platform in Embodiment 1 to make it more practical.

[0119] This application comprehensively considers the user's identity information, interests, hobbies, social behavior and other multi-dimensional characteristics to construct a detailed and accurate user portrait, and makes matching recommendations based on the similarity of the user feature models between the users to be matched, avoiding the limitations of relying solely on single-dimensional information for matching, greatly improving the accuracy and reliability of user matching in traditional social platforms, and providing users with social matching recommendations that are more in line with their personalized needs, significantly improving user satisfaction and engagement.

[0120] Those skilled in the art will appreciate that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required for implementing the present invention. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario may be distributed in the devices in the implementation scenario according to the description of the implementation scenario, or may be changed accordingly and located in one or more devices different from the present implementation scenario.

Claims

1. A user matching method based on a social platform, characterized in that: include: Acquire multi-dimensional user information data from one or more preset data sources, and pre-process the user information data; Extracting features from the preprocessed user information data based on a preset feature extraction rule to generate user feature data, and assigning weights to the user feature data based on a preset weighting method; Constructing a user feature model for the user according to the weight of each of the user feature data and a preset model training method; The similarity of the user feature model between the user to be matched and other users is calculated, and other users corresponding to the similarity reaching a preset value are recommended to the user to be matched based on a preset recommendation strategy.

2. The method according to claim 1, characterized in that: The user information data includes: identity information data, and / or interest data, and / or online research data, and / or social platform interaction data.

3. The method according to claim 1, characterized in that The feature extraction of the preprocessed user information data further includes: The data type of the user information data is determined, and different feature extraction methods correspond to different data types of the user information data.

4. The method according to claim 1, characterized in that The preset weighting method includes: subjective weighting method, and / or objective weighting method, and / or combined weighting method.

5. The method according to claim 1, characterized in that The “building a user feature model for the user according to the weight of each user feature data and a preset model training method” specifically includes: Preliminarily constructing a user feature model for the user according to the weight of each of the user feature data; Determining the existing social relationship data of the user, and using the social relationship data as a training sample for the preliminarily constructed user feature model; The parameters of the user feature model are adjusted by optimizing an objective function based on the training samples to determine a final user feature model for the user.

6. The method according to claim 1, characterized in that The “recommending other users corresponding to the similarity reaching a preset value to the to-be-matched user based on a preset recommendation strategy” also includes: Sorting the similarities between the to-be-matched user and each of the other users; The other users corresponding to the similarities that meet a preset threshold are sequentially formed into a recommendation list, and the recommendation list is sent to the user to be matched, wherein the recommendation list includes the user information data corresponding to each of the other users.

7. The method according to claim 1, characterized in that Also includes: Monitoring the user characteristic data of the user at a preset frequency, and if the user characteristic data changes, dynamically adjusting the weight of each user characteristic data and the user characteristic model for the user according to the change; The similarity between the user characteristic model adjusted by the user and the user characteristic models of other users is recalculated, and other users corresponding to the similarities reaching a preset value are recommended to the user based on a preset recommendation strategy.

8. The method according to claim 1, characterized in that Also includes: One or any multiple of the user information data, the user feature data, and the user feature model are associated with the user and then stored.

9. A user matching system based on a social platform, characterized in that: The method for implementing the user matching method based on a social platform according to any one of claims 1 to 8 comprises: An acquisition module, used to acquire multi-dimensional user information data from one or more preset data sources, and pre-process the user information data; An extraction module, used for performing feature extraction on the preprocessed user information data based on a preset feature extraction rule to generate user feature data, and assigning weights to the user feature data based on a preset weighting method; A construction module, used to construct a user feature model for the user according to the weight of each user feature data and a preset model training method; The recommendation module is used to calculate the similarity of the user feature model between the user to be matched and other users, and recommend other users corresponding to the similarity reaching a preset value to the user to be matched based on a preset recommendation strategy.

10. The system according to claim 9, characterized in that The feature extraction of the preprocessed user information data further includes: The determination module is used to determine the data type of the user information data, and the user information data of different data types correspond to different feature extraction methods.

Images